Repository navigation
v0.3.3 — Sound Mode, Input Type & Output Channels
Complete signal-chain visibility
Three new diagnostic sensors that finish the "what is my AVR actually doing right now" picture, alongside the format info added in v0.3.0/v0.3.1.
New entities
| Entity | Source | Example |
|---|---|---|
sensor.denon_sound_mode |
telnet MS? |
STEREO, MOVIE, PURE DIRECT, DOLBY ATMOS |
sensor.denon_input_signal_type |
telnet SD? |
eARC, HDMI, Analog, Digital (Coax/Opt) |
sensor.denon_output_channels |
derived | 2.1, 5.1, 7.1.4 |
Now you can see the full chain:
🎵 Dolby Audio - DD · 48 kHz
🔊 Mode: STEREO → 2.1
🔌 Input: eARC
📺 4K50 → 4K50 · Passthrough
About Output Channels
Because the AVR does not expose the active channel layout directly on X3700H firmware, this sensor is an educated guess derived from sound_mode + SYSDA input format. It's accurate for typical content (Stereo → 2.1, Movie decoding Atmos → 7.1.4, etc.), but exotic upmix cases may differ. The sensor's attributes expose source_channels, sound_mode, and is_derived: true so this is always transparent.
Upgrade
- HACS users: update → v0.3.3 → restart Home Assistant.
- Manual users: replace
custom_components/denon_advanced_audio/files and restart.
Existing entities keep their unique IDs.
Full changelog
- ✨ Add
Sound Modesensor (telnetMS?) - ✨ Add
Input Signal Typesensor (telnetSD?) - ✨ Add
Output Channelssensor (derived from mode + input) - 🔌
telnet.pyextended with new command queries andSD_MAPlabel table - 🔧
api.pyextended to route the new fields (idempotent patch) - 🏷️ Version bumped to
0.3.3
Full diff: v0.3.2...v0.3.3